Antitrust - Faculty Bios | Berkeley Haasfaculty.haas.berkeley.edu/shapiro/antitrust2007.pdfindustrial organization economics that are relevant for antitrust law. That task is far too

Chapter 15

ANTITRUST

LOUIS KAPLOW

School of Law, Harvard University, and National Bureau of Economic Research

CARL SHAPIRO

Haas School of Business and Department of Economics, University of California, Berkeley

Contents

1. Introduction 10772. Market power 1078

2.1. Definition of market power 10792.2. Single-firm pricing model accounting for rivals 10802.3. Multiple-firm models 1083

2.3.1. Cournot model with homogeneous products 10832.3.2. Bertrand model with differentiated products 10852.3.3. Other game-theoretic models and collusion 1086

2.4. Means of inferring market power 10872.4.1. Price-cost margin 10872.4.2. Firm’s elasticity of demand 10902.4.3. Conduct 1094

2.5. Market power in antitrust law 10953. Collusion 1098

3.1. Economic and legal approaches: an introduction 10993.1.1. Economic approach 10993.1.2. Legal approach 1101

3.2. Oligopoly theory 11033.2.1. Elements of successful collusion 11033.2.2. Repeated oligopoly games and the folk theorem 11043.2.3. Role of communications 1106

3.3. Industry conditions bearing on the likelihood of collusive outcomes 11083.3.1. Limited growth for defecting firm 11083.3.2. Imperfect detection 11093.3.3. Credibility of punishment 11103.3.4. Market structure 1112

Handbook of Law and Economics, Volume 2Edited by A. Mitchell Polinsky and Steven Shavell© 2007 Elsevier B.V. All rights reservedDOI: 10.1016/S1574-0730(07)02015-4

http://dx.doi.org/10.1016/S1574-0730(07)02015-4

1074 L. Kaplow and C. Shapiro

3.3.5. Product differentiation 11173.3.6. Capacity constraints, excess capacity, and investment in capacity 11173.3.7. Market dynamics 1119

3.4. Agreements under antitrust law 11213.4.1. On the meaning of agreement 11213.4.2. Agreement, economics of collusion, and communications 1124

3.5. Other horizontal arrangements 11293.5.1. Facilitating practices 11303.5.2. Rule of reason 1132

3.6. Antitrust enforcement 11363.6.1. Impact of antitrust enforcement on oligopolistic behavior 11373.6.2. Determinants of the effectiveness of antitrust enforcement 1137

4. Horizontal mergers 11384.1. Oligopoly theory and unilateral competitive effects 1139

4.1.1. Cournot model with homogeneous products 11394.1.2. Bertrand model with differentiated products 11434.1.3. Bidding models 1148

4.2. Oligopoly theory and coordinated effects 11494.3. Empirical evidence on the effects of horizontal mergers 1152

4.3.1. Stock market prices 11534.3.2. Accounting measures of firm performance 11544.3.3. Case studies 1155

4.4. Antitrust law on horizontal mergers 11574.4.1. Background and procedure 11574.4.2. Anticompetitive effects 11604.4.3. Efficiencies 1162

4.5. Market analysis under the Horizontal Merger Guidelines 11694.5.1. Market definition: general approach and product market definition 11704.5.2. Geographic market definition 11754.5.3. Rivals’ supply response 1177

4.6. Predicting the effects of mergers 11784.6.1. Direct evidence from natural experiments 11784.6.2. Merger simulation 1179

5. Monopolization 11805.1. Monopoly power: economic approach 1181

5.1.1. Rationale for monopoly power requirement 11815.1.2. Application to challenged practices 1183

5.2. Legal approach to monopolization 11865.2.1. Monopoly power 11875.2.2. Exclusionary practices 1191

5.3. Predatory pricing 11945.3.1. Economic theory 11955.3.2. Empirical evidence 1196

Ch. 15: Antitrust 1075

5.3.3. Legal test 11985.4. Exclusive dealing 1203

5.4.1. Anticompetitive effects 12035.4.2. Efficiencies 12095.4.3. Legal test 1210

6. Conclusion 1213Acknowledgements 1214References 1214Cases 1224


Abstract

This is a survey of the economic principles that underlie antitrust law and how thoseprinciples relate to competition policy. We address four core subject areas: marketpower, collusion, mergers between competitors, and monopolization. In each area, weselect the most relevant portions of current economic knowledge and use that knowl-edge to critically assess central features of antitrust policy. Our objective is to foster theimprovement of legal regimes and also to identify topics where further analytical andempirical exploration would be useful.

Keywords

antitrust, competition policy, monopoly, market power, market definition, oligopoly,collusion, cartels, price-fixing, facilitating practices, mergers, horizontal mergers,unilateral effects, coordinated effects, monopolization, exclusionary practices,predatory pricing, exclusive dealing

JEL classification: K21, L12, L13, L40, L41, L42


1. Introduction

In this chapter, we survey the economic principles that underlie antitrust law and usethese principles to illuminate the central challenges in formulating and applying compe-tition policy. Our twin goals are to inform readers about the current state of knowledgein economics that is most relevant for understanding antitrust law and policy and tocritically appraise prevailing legal principles in light of current economic analysis.

Since the passage of the Sherman Act in 1890, antitrust law has always revolvedaround the core economic concepts of competition and market power. For over a cen-tury, it has been illegal in the United States for competitors to enter into price-fixingcartels and related schemes and for a monopolist to use its market power to stifle com-petition. In interpreting the antitrust statutes, which speak in very general terms, U.S.courts have always paid attention to economics. Yet the role of economics in shapingantitrust law has evolved greatly, especially over the past few decades. The growing in-fluence of economics on antitrust law can be traced in part to the Chicago School, which,starting in the 1950s, launched a powerful attack on many antitrust rules and case out-comes that seemed to lack solid economic underpinnings. But the growing influence ofeconomics on antitrust law also has resulted from substantial theoretical and empiricaladvances in industrial organization economics over the period since then. With a lag,often spanning a couple of decades, economic knowledge shapes antitrust law. It is ourhope in this essay both to sharpen economists’ research agendas by identifying openquestions and difficulties in applying economics to antitrust law, and also to acceleratethe dissemination of economic knowledge into antitrust policy.

Antitrust economics is a broad area, overlapping to a great extent with the field of in-dustrial organization. We do not offer a comprehensive examination of the areas withinindustrial organization economics that are relevant for antitrust law. That task is fartoo daunting for a single survey and is already accomplished in the form of the three-volume Handbook of Industrial Organization (1989a, 1989b, 2007).1 Instead, we focusour attention on four core economic topics in antitrust: the concept of market power(section 2), the forces that facilitate or impede efforts by competitors to engage in collu-sion (section 3), the effects of mergers between competitors (section 4), and some basicforms of single-firm conduct that can constitute illegal monopolization, namely preda-tory pricing and exclusive dealing (section 5).2 In each case, we attempt to select fromthe broad base of models and approaches the ones that seem most helpful in formulat-ing a workable competition policy. Furthermore, we use this analysis to scrutinize thecorresponding features of antitrust law, in some cases providing a firmer rationalization

1 Schmalensee and Willig (1989a, 1989b) and Armstrong and Porter (2007).2 Since the field of antitrust economics and law is far too large to cover in one chapter, we are forced to omit

some topics that are very important in practice and have themselves been subject to extensive study, includingjoint ventures (touched on briefly in subsection 3.5.2), vertical mergers, bundling and tying, vertical intrabrandrestraints, the intersection of antitrust law and intellectual property law, and most features of enforcementpolicy and administration, including international dimensions.


for current policy and in others identifying important divergences.3 For reasons of con-creteness and of our own expertise, we focus on antitrust law in the United States, butwe also emphasize central features that are pertinent to competition policy elsewhereand frequently relate our discussion to the prevailing regime in the European Union.4

2. Market power

The concept of market power is fundamental to antitrust economics and to the law. Ex-cept for conduct subject to per se treatment, antitrust violations typically require thegovernment or a private plaintiff to show that the defendant created, enhanced, or ex-tended in time its market power. Although the requisite degree of existing or increasedmarket power varies by context, the nature of the inquiry is, for the most part, qualita-tively the same.

It is important to emphasize at the outset that the mere possession of market poweris not a violation of antitrust law in the United States. Rather, the inquiry into marketpower is usually a threshold question; if sufficient market power is established, it isthen asked whether the conduct in question—say, a horizontal merger or an alleged actof monopolization—constitutes an antitrust violation. If sufficient market power is notdemonstrated, the inquiry terminates with a victory for the defendant.

Here, we begin our treatment of antitrust law and economics with a discussion ofthe basic economic concept of market power and its measurement. Initially, we definemarket power, emphasizing that, as a technical matter, market power is a question of de-gree. Then we explore the factors that determine the extent of market power, first whenexercised by a single firm and then in the case in which multiple firms interact. We alsoconsider various methods of inferring market power in practice and offer some furtherremarks about the relationship between the concept of market power as understood by

3 There are a number of books that have overlapping purposes, including Bork (1978), Hylton (2003), Posner(2001), and Whinston (2006), the latter being closest to the present essay in the weight given to formaleconomics.4 As implied by the discussion in the text, our references to the law are primarily meant to make con-

crete the application of economic principles (and secondarily to offer specific illustrations) rather thanto provide detailed, definitive treatments. On U.S. law, the interested reader should consult the exten-sive treatise Antitrust Law by Areeda and Hovenkamp, many volumes of which are cited throughoutthis essay. On the law in the European Union, see, for example, Bellamy and Child (2001), Dabbah(2004), and Walle de Ghelcke and Gerven (2004). A wide range of additional information, includingformal policy statements and enforcement statistics, are now available on the Internet. Helpful linksare: Antitrust Division, Department of Justice: http://www.usdoj.gov/atr/index.html; Bureau of Competi-tion, Federal Trade Commission: http://www.ftc.gov/ftc/antitrust.htm; European Union, DG Competition:http://ec.europa.eu/comm/competition/index_en.html; Antitrust Section of the American Bar Association:http://www.abanet.org/antitrust/home.html.

http://www.usdoj.gov/atr/index.htmlhttp://www.ftc.gov/ftc/antitrust.htmhttp://ec.europa.eu/comm/competition/index_en.htmlhttp://www.abanet.org/antitrust/home.html


economists and as employed in antitrust law.5 Further elaboration appears in sections 4and 5 on horizontal mergers and monopolization, respectively.

2.1. Definition of market power

Microeconomics textbooks distinguish between a price-taking firm and a firm with somepower over price, that is, with some market power. This distinction relates to the demandcurve facing the firm in question. Introducing our standard notation for a single firmselling a single product, we write P for the price the firm receives for its product, X forthe firm’s output, and X(P ) for the demand curve the firm perceives that it is facing,with X′(P ) ≤ 0.6 When convenient, we will use the inverse demand curve, P(X).A price-taking firm has no control over price: P(X) = P regardless of X, over somerelevant range of the firm’s output. In contrast, a firm with power over price can causeprice to rise or fall by decreasing or increasing its output: P ′(X) < 0 in the relevantrange. We say that a firm has “technical market power” if it faces a downward sloping(rather than horizontal) demand curve.

In practice almost all firms have some degree of technical market power. Although thenotion of a perfectly competitive market is extremely useful as a theoretical construct,most real-world markets depart at least somewhat from this ideal. An important reasonfor this phenomenon is that marginal cost is often below average cost, most notablyfor products with high fixed costs and few or no capacity constraints, such as computersoftware, books, music, and movies. In such cases, price must exceed marginal cost forfirms to remain viable in the long run.7 Although in theory society could mandate thatall prices equal marginal cost and provide subsidies where appropriate, this degree ofregulation is generally regarded to be infeasible, and in most industries any attemptsto do so are believed to be inferior to reliance upon decentralized market interactions.Antitrust law has the more modest but, it is hoped, achievable objective of enforcingcompetition to the extent feasible. Given the near ubiquity of some degree of technicalmarket power, the impossibility of eliminating it entirely, and the inevitable costs ofantitrust intervention, the mere fact that a firm enjoys some technical market power isnot very informative or useful in antitrust law.

5 Prior discussions of the general relationship between the economic conception of market power and itsuse in antitrust law include Areeda, Kaplow, and Edlin (2004, pp. 483–499), Kaplow (1982), and Landes andPosner (1981). For a recent overview, see American Bar Association, Section of Antitrust Law (2005).6 For simplicity, unless we indicate otherwise, we assume throughout this chapter that each firm sells a

single product. While this assumption is almost always false, in many cases it amounts to looking at a firm’soperations product-by-product. Obviously, a multi-product firm might have market power with respect to oneproduct but not others. When interactions between the different products sold by a multi-product firm areimportant, notably, when the firm sells a line of products that are substitutes or complements for each other,the analysis will need to be modified.7 Edward Chamberlin (1933) and Joan Robinson (1933) are classic references for the idea that firms in

markets with low entry barriers but differentiated products have technical market power.


Nonetheless, the technical, textbook notion of market power has the considerable ad-vantage that it is amenable to precise measurement, which makes it possible to identifypractices that enhance a firm’s power to a substantial degree. The standard measureof a firm’s technical market power is based on the difference between the price thefirm charges and the firm’s marginal cost. In the standard theory of monopoly pric-ing, a firm sets the price for its product to maximize profits. Profits are given byπ = PX(P ) − C(X(P )), where C(X) is the firm’s cost function. Differentiating withrespect to price, we get the standard expression governing pricing by a single-productfirm,

(1)P − MC

P= 1|εF | ,

where MC is the firm’s marginal cost, C′(X), and εF ≡ dXdP PX is the elasticity of demandfacing that firm, the “firm-specific elasticity of demand.”8 The left-hand side of thisexpression is the Lerner Index, the percentage gap between price and marginal cost,which is a natural measure of a firm’s technical market power:

m ≡ P − MCP

.

As noted earlier, some degree of technical market power is necessary for firms to covertheir costs in the presence of economies of scale. For example, if costs are given byC(X) = F + CX, then profits are given by π = PX − CX − F and the condition thatprofits are non-negative can be written as m ≥ F/PX, that is, the Lerner Index must beat least as large as the ratio of the fixed costs, F , to the firm’s revenues, R ≡ PX.

Before proceeding with our analysis, we note that, although anticompetitive harm cancome in the form of reduced product quality, retarded innovation, or reduced productvariety, our discussion will follow much of the economics literature and most antitrustanalysis in focusing on consumer harm that comes in the form of higher prices. Thislimitation is not as serious as may first appear because higher prices can serve as a looseproxy for other forms of harm to consumers.

2.2. Single-firm pricing model accounting for rivals

To aid understanding, we present a basic but flexible model showing how underlyingsupply and demand conditions determine the elasticity of demand facing a given firm.This model allows us to begin identifying the factors that govern the degree of technicalmarket power enjoyed by a firm. We also note that this same model will prove veryuseful conceptually when we explore below the impact of various practices on price.Studying the effects of various practices on price requires some theory of how firms set

8 Strictly speaking, the elasticity of demand facing the firm is endogenous, except in the special case ofconstant elasticity of demand, since it varies with price, an endogenous variable. All the usual formulas referto the elasticity of demand at the equilibrium (profit-maximizing) price level.


their prices. The building block for these various theories is the basic model of price-setting by a single, profit-maximizing firm. In addition, as a matter of logic, one mustbegin with such a model before moving on to theories that involve strategic interactionsamong rival firms.

The standard model involves a dominant firm facing a competitive fringe.9 A profit-maximizing firm sets its price accounting for the responses it expects from its rivals andcustomers to the price it sets.10 This is a decision-theoretic model, not a game-theoreticmodel, so it does not make endogenous the behavior of the other firms in the marketor of potential entrants. This is the primary sense in which the generality of the modelis limited. The model also is limited because it assumes that all firms in the marketproduce the same, homogeneous product and do not engage in any price discrimination,although the core ideas underlying it extend to models of differentiated products.

The firm faces one or more rivals that, as noted, sell the same, homogeneous product.When setting its price, P , the firm recognizes that rivals will likely respond to higherprices by producing more output. The combined output of the firm’s rivals increaseswith price according to Y(P ), with Y ′(P ) ≥ 0. Total (market) demand declines withprice according to Z(P ), with Z′(P ) ≤ 0. If the firm in question sets the price P , thenit will be able to sell an amount given by X(P ) ≡ Z(P ) − Y(P ). This is the largestquantity that the firm can sell without driving price below the level P that it selected; ifthe firm wants to sell more, it will have to lower its price. The firm’s so-called “residualdemand curve” is therefore given by X(P ).

If we differentiate the equation defining X(P ) with respect to P , and then multiplyboth sides by −P/X to convert the left-hand side into elasticity form, we get

−PX

dX

dP= −P

X

dZ

dP+ P

X

dY

dP.

Next, multiply and divide the dZ/dP term on the right-hand side by Z and the dY/dPterm by Y . This gives

−PX

dX

dP= −P

Z

dZ

dP

Z

X+ P

Y

dY

dP

Y

X.

Define the market share of the firm being studied by S = X/Z. The correspondingmarket share of the rivals is 1−S = Y/Z. Replacing Z/X by 1/S and Y/X by (1−S)/Sin the expression above gives

−PX

dX

dP= −P

Z

dZ

dP

1

S+ P

Y

dY

dP

(1 − S)S

.

9 For a recent textbook treatment of this model, see Carlton and Perloff (2005, pp. 110–119). Landes andPosner (1981) provide a nice exposition of this model in the antitrust context.10 As with the standard theory of pure monopoly pricing as taught in microeconomics textbooks, the resultsof this model are unchanged if we model the firm as choosing its output level, with price adjusting to clearthe market.


Call the elasticity of supply of the rivals εR ≡ PY dYdP , and the absolute value of theelasticity of the underlying market demand curve |εD| ≡ −PZ dZdP . The absolute value ofthe elasticity of demand facing the firm, |εF | ≡ −PX dXdP , is therefore given by

(2)|εF | = |εD| + (1 − S)εRS

.

This equation captures the central lesson from this model: the absolute value of theelasticity of demand facing a single firm, given the supply curves of its price-takingrivals and the demand curve of the buyers in its market, is governed by three variables:(1) the underlying elasticity of demand for the product, |εD|, which is frequently calledthe market elasticity of demand; (2) the elasticity of supply of the firm’s rivals, εR; and(3) the firm’s market share, S. The magnitude of the firm-specific elasticity of demand islarger, the larger are the magnitudes of the market elasticity of demand and the elasticityof supply of the firm’s rivals and the smaller is the firm’s market share. Intuitively,market share is relevant for two reasons: the smaller the firm’s share, the greater theshare of its rivals and thus the greater is the absolute magnitude of their supply responseto a price increase for a given supply elasticity, εR; and the smaller the firm’s share, thesmaller is its share of the increase in industry profits due to a given sacrifice in its ownsales.11

One polar case in this basic model is that of the traditional monopolist. With no ri-vals, S = 1, so the elasticity of demand facing the firm is just the market elasticity ofdemand. With rivals, however, the magnitude of the firm-specific elasticity of demandis larger than that of the market elasticity of demand. The other polar case is that of thefirm from the theory of perfectly competitive markets. As the firm’s share of the mar-ket approaches zero, the magnitude of the firm-specific elasticity of demand becomesinfinite, that is, the firm is a price-taker.

We can directly translate the firm-specific elasticity of demand given by expres-sion (2) into the profit-maximizing price. As indicated in expression (1), profit maxi-mization involves setting price so that the firm’s gross margin, m, equals the inverse ofthe magnitude of the firm’s elasticity of demand. If there are no rivals, S = 1 and thisrelationship simplifies to the standard monopoly formula, m = 1/|εD|. For a firm witha tiny market share, |εF | is enormous, so m ≈ 0, that is, price nearly equals marginalcost. For intermediate cases, as noted, in this model a large market elasticity of demand,|εD|, a high elasticity of rival supply, εR , and a small market share, S, all lead to a largefirm-specific elasticity of demand facing the price leader, |εF |, which in turn implies asmall margin.

This model provides a guide for studying the types of conduct that may enhancea firm’s technical market power and thus allow that firm profitably to raise its price.

11 It should be noted that statements about the effect of market share must be interpreted carefully. Thus, anoutward shift in the supply curve of the rivals, which lowers the firm’s market share at any given price, willraise the elasticity of demand facing that firm at any given price. However, more broadly, the firm’s marketshare is endogenous because it depends on the price the firm chooses.


Generically, such conduct will be that which reduces the value of the right side of ex-pression (2): conduct that makes substitute products less attractive, that causes rivalsto reduce their supply, and that raises the firm’s market share (through the two formermeans or otherwise). Later we consider how certain types of conduct having these ef-fects should be scrutinized under antitrust law.

This model is quite broad when one undertakes appropriate interpretations and exten-sions. For example, issues relating to substitute products bear on the market elasticityof demand, as will be noted below. Additionally, one can account for entry by reflectingit in the rival supply elasticity. One particular variant of the model involves infinitelyelastic rival supply, perhaps due to entry, at some fixed “limit” price.

2.3. Multiple-firm models

The model in subsection 2.2 took the behavior of all but one firm as exogenous. In thissection, we consider game-theoretic models that make predictions regarding the degreeof market power exercised by interacting firms. First we consider two standard, static,noncooperative models: Cournot’s model of oligopoly, for the case with homogeneousproducts, and Bertrand’s model, for the case with differentiated products. Then we con-sider briefly the possibility of repeated games and the impact of collusive behavior onmarket power.12

2.3.1. Cournot model with homogeneous products

The Cournot (1838) model of oligopoly with homogeneous products is similar to thesingle-firm pricing model in that it identifies how certain observable characteristics ofthe market determine the degree of a firm’s market power, that is, the percentage markupabove marginal cost that the firm charges. The Cournot model goes further, however, byproviding predictions about how market structure affects the equilibrium price, predic-tions that will be important for seeing how certain commercial practices and mergersaffect price. Specifically, the model predicts that firms with lower costs will have highermarket shares and higher markups. The model is frequently employed in markets withrelatively homogeneous products, especially if firms pick their output or capacity levels,after which prices are determined such that the resulting supply equals demand.13 How-ever, one should bear in mind that the Cournot equilibrium is the Nash equilibrium in aone-shot game. As we discuss at length in section 3, many different outcomes can ariseas equilibria in a repeated oligopoly game, even if the stage game played each periodinvolves quantity-setting à la Cournot. In antitrust applications, it is generally desirable

12 There is an enormous literature on oligopoly theory, which we do not attempt to cover systematically. See,for example, Shapiro (1989), Tirole (1988), and Vives (2001). We discuss models of repeated oligopoly atgreater length in section 3 on collusion.13 Kreps and Scheinkman (1983) use a particular rationing rule to show that capacity choices followed bypricing competition can replicate the Cournot equilibrium.


to test robustness of results to alternative solution concepts as well as to test empiricallythe predictions of any oligopoly model that is employed.

In a Cournot equilibrium, a single firm’s reaction curve is derived as a special caseof the basic model of single-firm pricing: the rivals’ outputs are all taken to be fixed,so the rival supply elasticity is zero. As we now show, the elasticity of demand facinga single firm is equal to the market elasticity of demand divided by that firm’s marketshare. However, the Cournot model goes beyond the single-firm pricing model becauseit involves finding the equilibrium in a game among multiple firms.

Suppose that there are N firms, with each firm i choosing its output Xi simul-taneously. The Cournot equilibrium is a Nash equilibrium in these quantities. Totaloutput is X ≡ X1 + · · · + XN . Industry or market (inverse) demand is given byP = P(X). Given the output of the other firms, firm i chooses its output to maxi-mize its own profits, πi = P(X)Xi − Ci(Xi). The first-order condition for this firm isP(X) + XiP ′(X) − C′i (Xi) = 0. This can be written as

(3)P − MCi

P= Si|εD| ,

where Si ≡ Xi/X is firm i’s market share, and εD , as before, is the market elasticity ofdemand.

To explore this result, consider the special case in which each firm i has constantmarginal cost MCi . Adding up the first-order conditions for all of the firms givesNP(X) + XP ′(X) = ∑i MCi , which tells us that total output and hence the equilib-rium price depend only upon the sum of the firms’ marginal costs. Moreover, the markupequation tells us that lower-cost firms have higher market shares and enjoy more tech-nical market power. At the same time, the larger is the market elasticity of demand forthis homogeneous product, the smaller is the market power enjoyed by each firm and thelower are the margins at all firms. Here we see a recurrent theme in antitrust: a lower-cost firm may well enjoy some technical market power and capture a large share of themarket, but this is not necessarily inefficient. Indeed, with constant marginal costs, fullproductive efficiency would call for the firm with the lowest marginal cost to serve theentire market.

The Cournot model also predicts that total output will be less than would be ef-ficient because none of the firms produces up to the point at which marginal costequals price; they all have some degree of market power. In the special case with con-stant and equal marginal costs, each firm has a market share of 1/N , and the modelpredicts that each enjoys technical market power according to the resulting equation(P − MC)/P = 1/N |εD|. In this simple sense, more firms leads to greater competi-tion and lower prices. However, this model is clearly incomplete for antitrust purposes:presumably, there are fixed costs to be covered (which is why there is a fixed numberof firms in the first place), so adding more firms is not costless.14 This type of analy-

14 In general, there is no reason to believe that the equilibrium number of firms in an oligopoly with freeentry, that is, where equally efficient firms enter until further entry would drive profits below zero, is socially


sis will be directly relevant when we consider horizontal mergers, which remove anindependent competitor but may also lead to efficiencies of various types.

One of the attractive theoretical features of the Cournot model is that it gener-ates an elegant formula for the industry-wide average, output-weighted, price-costmargin, that is, the expression PCM ≡ ∑Ni=1 Si P−MCiP . Using equation (3), we getPCM ≡ ∑Ni=1 Si Si|εD | or

(4)PCM = 1|εD|N∑

i=1S2i =

H

|εD| ,

where H ≡ ∑ S2i is the Herfindahl-Hirschman Index (HHI) of market concentrationthat is commonly used in antitrust analysis, especially of horizontal mergers.

2.3.2. Bertrand model with differentiated products

The Bertrand model with differentiated products is the other key static model ofoligopoly used in antitrust. The Bertrand equilibrium is the Nash equilibrium in thegame in which the firms simultaneously set their prices. With N firms selling differen-tiated products, we can write the demand for firm i’s product as Xi = Di(P1, . . . , PN).As usual, the profits of firm i are given by πi = PiXi − Ci(Xi). The Bertrand equi-librium is defined by the N equations ∂πi/∂Pi = 0. Writing the elasticity of demandfacing firm i as εi ≡ ∂Xi∂Pi PiXi , firm i’s first-order condition is the usual markup equation,

Pi − MCiPi

= 1|εi | .Actually solving for the Bertrand equilibrium can be difficult, depending on the func-tional form for the demand system and on the firms’ cost functions. In general, however,we know that a firm faces highly elastic demand if its rivals offer very close substitutes,so the Bertrand theory predicts larger markups when the products offered by the variousfirms are more highly differentiated. In practice, notably, in the assessment of mergers,particular models of product differentiation are used, such as discrete choice modelswith random utilities, including logit and nested logit models, or models with linear de-mand or constant elasticities, as we discuss further in section 4 on horizontal mergers.

Here we illustrate the operation of the Bertrand model by explicitly solving a simple,symmetric, two-firm model with constant marginal costs and linear demand. Write thedemand curves as X1 = A−P1 +αP2 and X2 = A−P2 +αP1. Note that the parameterα measures the diversion ratio, that is, the fraction of sales lost by one firm, when it

efficient. See, for example, Mankiw and Whinston (1986). This observation is relevant in assessing certainantitrust policies: if the equilibrium number of firms is “naturally” too small, then exclusionary conduct onthe part of the incumbent oligopolists creates an additional social inefficiency. However, if the equilibriumnumber of firms is “naturally” excessive, different implications would follow.


raises its price, that are captured by the other firm (assuming that the other firm’s priceis fixed). The diversion ratio, α, will be important when we study horizontal mergersbelow.15

Call the marginal costs per unit MC1 and MC2, respectively, and assume that thereare no fixed costs. Then we have π1 = (P1 − MC1)(A − P1 + αP2). Differentiatingwith respect to P1 and setting this equal to zero, we get firm 1’s best-response curve,P1 = (A + αP2 + MC1)/2. Assuming cost symmetry as well, MC = MC1 = MC2,in the symmetric Bertrand equilibrium we must have P1 = P2 = PB so we getPB = A+MC2−α .

We can compare the Bertrand equilibrium price to the price charged by a single firmcontrolling both products. Such a firm would set P to maximize (P −MC)(A−P +αP ),which gives the monopoly price of PM = A+MC(1−α)2(1−α) . The percentage gap betweenthe monopoly price and the Bertrand price is given by PM−PB

PB= α2(1−α) PB−MCPB .16 This

expression tells us that the Bertrand equilibrium price is relatively close to the monopolyprice when the two products are rather poor substitutes, that is, when the diversion ratio,α, is low.

This formula will be highly relevant when studying the effect on price of a mergerbetween two suppliers of differentiated products. In that context, the formula measuresthe price increase associated with the merger, given the prices charged by other firms(and before accounting for efficiencies). The price increase will depend on the pre-merger margin, PB−MC

PB, and on the diversion ratio.

2.3.3. Other game-theoretic models and collusion

Both the Cournot and Bertrand models assume that firms engage in a one-shot nonco-operative game. An extensive literature on repeated games explores the possibility thatfirms may do better for themselves, supporting what are more colloquially describedas collusive outcomes, approaching or equaling the industry profit-maximizing price.As suggested by Stigler (1964) and refined in subsequent work, higher prices tend tobe sustainable when cheating can be rapidly detected and effectively punished. For ageneral discussion of models of collusion, see Jacquemin and Slade (1989) and Shapiro(1989).

The possibility that firms can support alternative equilibria featuring higher prices isimportant to antitrust analysis. First, it suggests that market power may be higher than isotherwise apparent. Second and more important, the possibility of collusion affects theantitrust analysis of other business conduct. For example, a horizontal merger may have

15 More generally, the diversion ratio from product i to substitute product j is defined as

αji = (dXj /dPi)/(−dXi/dPi). Converting this equation into elasticity form gives αji = εji|εi |XjXi

, where

εji = dXjdPiPiXj

is the cross-elasticity from product i to product j .16 The details of these calculations are available at http://faculty.haas.berkeley.edu/shapiro/unilateral.pdf.

http://faculty.haas.berkeley.edu/shapiro/unilateral.pdf


only a minor impact on price if the merging firms and their rivals are already colluding,but a far greater effect if the reduction in the number of competitors makes collusioneasier to sustain. Also, some practices may facilitate collusion, in which case such prac-tices themselves should potentially be subject to antitrust scrutiny. These possibilitiesare explored further in section 3 on collusion and section 4 on horizontal mergers.

2.4. Means of inferring market power

Assessing the extent of or increase in technical market power in a given situation is oftena difficult undertaking. Based upon the foregoing analysis, one can identify a numberof potential strategies whose usefulness varies greatly by context. The legal system hastended to rely primarily on a subset of these approaches, focusing mostly on marketdefinition, as discussed below. In recent years, however, it has increasingly consideredalternatives when it has perceived that credible economic evidence has been offered.17

Although somewhat crude, it is helpful to group means of inferring market powerinto three categories. First, since market power is technically defined by the extent ofthe price-cost margin, one can attempt to identify evidence that bears fairly directly onthe size of this margin, or to measure profits (which reflect the margin between priceand average cost). Second, various models, such as the single-firm price-setting modelin subsection 2.2, indicate that the extent of market power will be a function of theelasticity of demand, a firm’s market share, and rivals’ supply response. Accordingly,one can analyze information indicative of the magnitude of these factors. Third, one canmake inferences from firm behavior, notably when observed actions would be irrationalunless a certain degree of market power existed or was thereby conferred.

2.4.1. Price-cost margin

2.4.1.1. Direct measurement Observing the extent to which price is above marginalcost indicates the degree of technical market power. This direct approach is feasible ifone can accurately measure price and some version of marginal cost, usually averageincremental cost.18 Price is often easy to identify, although complications may arisewhen multiple products are sold together, making it difficult to determine the incremen-tal revenue associated with the product in question. If different customers are charged

17 For example, the Supreme Court in Federal Trade Commission v. Indiana Federation of Dentists, 476 U.S.447, 460–461 (1986) (quoting Areeda’s Antitrust Law treatise) stated: “Since the purpose of the inquiries intomarket definition and market power is to determine whether an arrangement has the potential for genuineadverse effects on competition, ‘proof of actual detrimental effects, such as a reduction of output,’ can obviatethe need for an inquiry into market power, which is but a ‘surrogate for detrimental effects.’ ”18 We use the terms “marginal cost” and “average incremental cost” interchangeably. Both measure the extracost per unit associated with increased output. Average incremental cost is a somewhat more accurate term,since one is often interested in increments that do not correspond to “one unit” of output. However, if onetakes a flexible approach to what constitutes a “unit” of production, the two terms are exactly the same. Inpractice, average incremental cost is used to determine gross profit margins.


different prices, it may be necessary to calculate the profit margins for sales to differ-ent customers (or at different points of time). Complexities also arise when some salesimplicitly bundle other services, such as delivery, short-term financing, and customersupport; in principle, these factors can be accommodated by redefining the product toinclude these services (and tracking the costs associated with these services). Marginalcost, by contrast, may be more difficult to measure, due both to difficulties in identifyingwhich costs are variable (and over what time period) and to the presence of commoncosts that may be difficult to allocate appropriately. In part for this reason, the empiricalindustrial organization literature, surveyed in Bresnahan (1989), often treats marginalcost as unobservable.

In some cases, approximate measures of price-cost margins may be sufficient andeasy to produce, but as evidenced by disputes over cost in predatory pricing cases andin various regulatory contexts, direct measurement of any conception of cost can bedifficult and contentious. In any event, as with all measures of technical market power,it is important to keep in mind the distinction between the extent of market power andwhether particular conduct should give rise to antitrust liability. For example, as wehave already noted, especially in industries in which marginal cost is below averagecost and capacity constraints are not binding, nontrivial technical market power may beconsistent with what are normally considered competitive industries.

2.4.1.2. Price comparisons Another fairly direct route to assessing the magnitude ofprice-cost margins, or at least to provide a lower-bound estimate, is to compare pricesacross markets. For example, if a firm sells its product for a substantially higher pricein one region than in another (taking into account transportation and other cost differ-ences), the price-cost margin in the high-price region should be at least as great as the(adjusted) price difference between the regions. This inference presumes, of course, thatthe price in the low-price region is at least equal to marginal cost. Note that this methodcan be understood as a special case of direct measurement. It is assumed that the lowprice is a proxy for (at least an upper bound on) marginal cost, and one then is measuringthe price-cost margin directly.

The Staples merger case illustrates an application of this method.19 The governmentoffered (and the court was convinced by) data indicating that prices were higher inregional markets in which fewer office supply superstores operated and that prices fellwhen new superstore chains entered. This was taken as powerful evidence that a mergerof two of the existing three superstores would lead to price increases.

2.4.1.3. Price discrimination Price comparisons often involve a special case of pricediscrimination, wherein a given firm charges different prices to different consumers,contrary to the implicit assumption in the earlier analysis that each firm sets a single

19 Federal Trade Commission v. Staples, Inc., 970 F. Supp. 1066 (D.D.C. 1997). For further discussion, seesubsection 4.6.1 in our discussion of horizontal mergers.


price for all of its customers. Accordingly, for essentially the same reason as that justgiven, the ability of a firm to engage in price discrimination implies the existence ofmarket power. If one is prepared to assume that the firm is not pricing below marginalcost to any of its customers, and if one accounts for differences in the cost of servingdifferent customers, the percentage difference between any high price it charges andthe lowest price it charges for the same product can serve as a lower bound on thepercentage markup associated with the higher price. For example, the substantial pricediscrimination in sales of pharmaceutical drugs on international markets shows thatprices in the United States are very much above marginal cost.

The fact that price discrimination technically implies market power is important be-cause price discrimination is widespread. Familiar examples include airline pricing,senior citizen and student discounts, and the mundane practice of restaurants chargingsteep price increments for alcoholic beverages (versus soft drinks) and high-end entreesthat greatly exceed any differences in marginal cost. For business-to-business transac-tions, negotiations that typically generate price dispersion and price discrimination arequite common.

Once again, however, it is important to keep in mind that the existence of technicalmarket power does not imply antitrust liability.20 As is familiar, price discriminationgenerates greater seller profits yet may well be benign or even favorable on average forconsumers. Moreover, the resulting profit margins are often necessary to cover fixedcosts, as in models of monopolistic competition. If there are no barriers to entry so thatthe resulting margins merely provide a normal rate of return on capital, the presenceof a gap between price and marginal cost is perfectly consistent with the conclusionthat the market is behaving is a competitive fashion, given the presence of fixed costsand product differentiation. Furthermore, in our preceding example of multinationalpharmaceutical companies, the margins provide the reward for costly and risky researchand development to create and patent new drugs. The ex post market power is necessaryto provide the quasi-rents that induce innovation (given that we rely on a patent systemrather than a regime that gives direct rewards to innovators from the government fisc).

2.4.1.4. Persistent profits A somewhat different approach to establishing antitrustmarket power involves looking at a firm’s profits, which amounts to comparing priceto average (rather than marginal) cost. Under this approach, persistently above-normal

20 Nor is it the case that price discrimination in itself implies antitrust liability, despite the existence of theRobinson-Patman Act that regulates particular sorts of price discrimination in certain contexts. As presentlyinterpreted, price discrimination may be a violation in so-called primary-line cases, tantamount to predatorypricing, and in secondary-line cases, such as when manufacturers offer discounts (that are not cost justified)to large retailers that are not available to smaller buyers. Notably, the Act does not cover discriminatory pricesto ultimate consumers (or to intermediaries that are not in competition with each other) that are nonpredatory.Nevertheless, it seems that defendants in antitrust litigation have been reluctant to rationalize challenged prac-tices that analysts have suggested were means of price discrimination on such grounds, presumably fearingthat such explanations would be to their detriment. Of course, one way this could be true is that the existenceof some technical market power would thereby be conceded.


profits indicate a high price-cost margin and thus the existence of technical marketpower. This method shares difficulties with any that rely on measures of cost. In partic-ular, it is often very hard to measure the return on capital earned for a given product, orin a given market, especially for a firm that is engaged in many lines of business andhas substantial costs that are common across products.21 Another problem with this ap-proach is that the return on capital should, in principle, be adjusted for risk. Frequently,one is looking at a successful firm, perhaps one that has been highly profitable for manyyears following some initial innovation that, ex ante, may not have turned out as well.

In addition, average costs often differ from marginal costs. When average costs arehigher, this approach may mask the existence of technical market power. In such cir-cumstances, however, marginal-cost pricing may be unsustainable in any event; that is,although there may be technical market power, there may not be any way (short of in-trusive regulation that is not contemplated) to improve the situation. When average costis below marginal cost, profits can exist despite the absence of any markup. In suchcases, entry might be expected. If profits are nevertheless persistent, there may existentry barriers, a subject we discuss below.

2.4.2. Firm’s elasticity of demand

In the single-firm pricing model, the price-cost margin (Lerner Index) equals the inverseof the (absolute value of the) firm’s elasticity of demand, as indicated by expression (1).Furthermore, as described in expression (2), this elasticity depends on the market elas-ticity of demand, the firm’s market share, and rivals’ supply elasticity. In the Cournot,Bertrand, and other oligopoly models, many of the same factors bear on the extent ofthe price-cost margin and thus the degree of market power. Accordingly, another routeto inferring market power is to consider the magnitude of these factors.

2.4.2.1. Direct measurement One could attempt to measure the elasticity of demandfacing the firm in question.22 A possible approach would be to estimate the marketelasticity of demand and then make an adjustment based on the firm’s market share.Alternatively, one might directly observe how the firm’s sales have varied when it haschanged its price. As a practical matter, both of these methods may be difficult to im-plement. However, they may nevertheless be more reliable than the alternatives.

2.4.2.2. Substitutes, market definition, and market share In antitrust analysis, both byagencies (notably, in examining prospective horizontal mergers) and by the courts, thedominant method of gauging the extent of market power involves defining a so-calledrelevant market and examining the share of a firm or group of firms in that market. Indefining product markets, the focus is on which products are sufficiently good demand

21 See, for example, Fisher and McGowan (1983).22 See, for example, Baker and Bresnahan (1988).


substitutes for the product in question to be deemed in the same market. Likewise, indefining the extent of the geographic market, the question concerns the feasibility ofsubstitution, for example, by asking how far patients would travel for hospitalization.Although we have discussed the economic analysis of market power at some length, theconcept of market definition has not yet appeared directly. Hence it is useful to considerthe relationship between the most common method used in antitrust law to assess marketpower and the implications of the foregoing economic analysis.

The connection is easiest to see by examining expression (2), which relates the firm-specific elasticity of demand to the market elasticity of demand, the firm’s market share,and rivals’ elasticity of substitution. Consider the case in which the firm produces a ho-mogeneous product, has a high share of sales of that product, and faces a highly elasticmarket demand curve due to the existence of many close substitutes. The firm-specificelasticity of demand is high and thus the extent of technical market power is small eventhough the firm’s market share is high in the narrowly defined market consisting only ofthe homogeneous product sold by the firm. One could redefine the “market” to includethe close substitutes along with the homogeneous product sold by the firm. The marketelasticity of demand in this broader market is presumably smaller, but since the firm’smarket share in this market is also necessarily lower, we would again conclude thatthe firm-specific demand elasticity is large and thus that the degree of technical marketpower is low.

Courts—and thus lawyers and government agencies—traditionally equate high mar-ket shares with a high degree of market power and low shares with a low degree ofmarket power. This association is highly misleading if the market elasticity of demandis ignored, and likewise if rivals’ elasticity of supply is not considered. In principle,as just explained, the paradigm based on market definition and market share takes themarket elasticity of demand into account, indirectly, by defining broader markets—andthus producing lower market shares—when the elasticity is high. As should be apparentfrom the foregoing discussion, the standard antitrust approach is more indirect than nec-essary and, due to this fact plus its dichotomous structure (substitutes are either in themarket or not), will tend to produce needlessly noisy conclusions.23 We discuss marketdefinition at greater length in subsection 4.5 on horizontal mergers and subsection 5.2.1on monopolization.

Frequently, it is useful to decompose the elasticity of demand for a given productinto various cross-elasticities of demand with other products. For example, if the priceof soda rises, consumers will substitute to other drinks, including, perhaps, beer, juice,milk, and water. Naturally, the analysis in any given case will depend upon exactlyhow these various products are defined (soda could be broken into regular soda anddiet soda, or colas and non-colas, etc.). But the underlying theory of demand does notvary with such definitions. To illustrate, suppose that consumers allocate their total in-come of I across N distinct products, so

∑Ni=1 PiXi = I . To study the elasticity of

23 This point is elaborated in Kaplow (1982).


demand for product 1, suppose that P1 rises and the other prices remain unchanged.Then we get X1 + P1 dX1dP1 +

∑Ni=2 Pi

dXidP1

= 0. Converting this to elasticity form gives− P1

X1

dX1dP1

= 1 + ∑Ni=2 P1XidXidP1

PiXiP1X1

. Defining the cross-elasticity between product i and

product 1 as εi1 = dXidP1 P1Xi , and the revenues associated with product i as Ri = PiXi ,this can be written as

(5)|ε11| = 1 +N∑

i=2εi1

Ri

R1.

In words, the (absolute value of the) elasticity of demand for product 1 is equal toone plus the sum of the cross-elasticities of all the other products with product 1, witheach cross-elasticity weighted by the associated product’s revenues relative to thoseof product 1. If we define each product’s share of expenditures as si = Ri/I , thenexpression (5) can be written as |ε11| = 1 + 1s1

∑Ni=2 siεi1, so the cross-elasticity with

each rival product is weighted by its share of revenues.24

This decomposition of the market elasticity of demand is instructive with regard tothe standard practice in antitrust of defining markets by deciding whether particularproducts are sufficiently good substitutes—generally understood as having sufficientlyhigh cross-elasticities of demand—to be included in the market. The expression makesclear that even a substitute with a very high cross-elasticity may have much less influ-ence than that of a large group of other products, no one of which has a particularlyhigh cross-elasticity. Moreover, products’ shares of total revenues are not ordinarilyconsidered in an explicit way, yet the formula indicates that a substitute with half thecross-elasticity of another can readily be more important, in particular, if its associatedrevenues are more than twice as high. More broadly, this representation of the rela-tionship between overall elasticity and individual cross-elasticities reinforces the pointthat the effect of substitutes is a matter of degree and thus not well captured by theall-or-nothing approach involved in defining antitrust markets.

Some further comments concerning market share are in order, particularly in lightof the fact that a persistently high market share is very frequently presented as com-pelling evidence that a firm has market power. No doubt this inference is often valid,specifically, if the market demand elasticity and rivals’ supply elasticities are low inmagnitude and the market conditions are reasonably stable. However, a firm with only amodest cost advantage may profitably maintain its high share by pricing low enough tocapture most of the market. This occurs, for example, in the model of the dominant firmfacing a competitive fringe if the fringe supply is very elastic at a price just above thefirm’s own marginal cost. Consider, for example, a trucking firm that provides 100% of

24 Cross-elasticities need not be positive. For example, when the weighted summation equals zero, we havethe familiar case of unit elasticity—that is, as price rises, expenditures on the product in question remainconstant—and when the summation is negative, we have an elasticity less than one in absolute value, oftenreferred to as inelastic demand.


the freight transportation on a particular route but would quickly be displaced by nearbyrivals (whose costs are essentially the same but who suffer a slight disadvantage due toa lack of familiarity with the route’s customers) if it were to raise its price even a fewpercent. Additionally, a firm may have a 100% share in a market protected by a patent,but if there are sufficiently close substitutes, its market power is negligible. Conversely,even a firm with a low share of sales of a particular product may have quite a bit oftechnical market power if the magnitude of the market elasticity of demand and rivals’elasticity of supply for that product are very low. Gasoline refining and electricity gen-eration are two examples of products for which this latter situation can arise. In sum, theright side of expression (2) indicates that market share is only one factor that determinesthe elasticity of demand facing a firm, so the magnitude of market share is a relevantcomponent of market power but not a conclusive indicator.

2.4.2.3. Rivals’ supply response: barriers to expansion, mobility, and entry In exam-ining the right side of expression (2) for the firm’s elasticity of demand, the precedingsubsection focused on the market elasticity of demand and market share. However, theelasticity of supply by rivals is also relevant, as indicated by the just-mentioned contrast-ing examples of trucking, on one hand, and gasoline refining and electricity generation,on the other hand. The concept of rivals’ supply should be understood broadly, to in-clude expanded output from existing plants, shifting capital from other regions or fromthe production of other products, introducing new brands or repositioning existing ones,and entry by firms in related businesses or by other firms. If market power is significant,it must be that the aggregate of these potential supply responses—often referred to asexpansion, mobility, and entry—is sufficiently limited, at least over some pertinent timeperiod. Gilbert (1989) provides an extended discussion of such barriers, Berry and Reiss(2007) survey empirical models of entry and market power, and Sutton (2007) discussesthe relationship between market structure and market power.

In some cases, the elasticity of rivals’ supply may be measured directly, by measuringoutput responses to previous changes in price by the firm in question, or by other firms insimilar markets. Often, however, some extrapolation is required, such as in predictingwhether a hypothetical increase in price to unprecedented levels following a mergerwould generate a significant supply response. For internal expansion by existing rivals,the question would be whether there exist capacity constraints, steeply rising marginalcosts, or limits on the inclination of consumers of differentiated products to switchallegiances. In the case of new entry, timing, possible legal restrictions (intellectualproperty, zoning, and other regulatory constraints), brand preferences, the importanceof learning by doing, and the ability to recoup fixed costs, among other factors, willdetermine the extent of restraint imposed.

Particularly regarding the latter, it is common to inquire into the existence of so-calledbarriers to entry (sometimes taken as a shorthand for all forms of supply response byrivals). In some instances, such as when there are legal restrictions, the meaning ofthis concept is fairly clear. However, in many cases, it is difficult to make sense of thenotion of entry barriers in a vacuum. For example, there is much debate about whether


economies of scale should be viewed as a barrier to entry. If minimum efficient scaleis large and incumbent producers have long-term exclusive dealing contracts with mostdistributors, entry may be rendered too costly, and existing firms might enjoy high price-cost margins (more than necessary to cover fixed costs). If instead there merely existfixed costs and marginal costs are constant, in a free-entry equilibrium there will bepositive price-cost margins yet no profits. The positive margins will not induce furtherentry because their level post-entry would be insufficient to recover fixed costs. As wehave observed repeatedly, although market power would exist in the technical sense, thesituation should not be viewed as problematic from an antitrust perspective.

Many structural features of markets have been identified as possible entry barriers:economies of scale, learning by doing, reputation, access to capital, customer switchingcosts, lack of product compatibility, network effects, patent protection, and access todistribution channels. Because the implication of so-called entry barriers depends onthe context—and because some degree of market power is sometimes unavoidable yetmany are reluctant to state or imply its existence, such as by deeming something tobe an entry barrier in a setting where antitrust intervention seems inappropriate—thereis no real consensus on how the term “barriers to entry” should be defined or appliedin practice.25 We do not see clear benefits to formulating a canonical definition of theconcept. It may be best simply to keep in mind the purpose of such inquiries into theexistence of entry barriers: to assess rivals’ supply response as an aspect of an inquiryinto the existence of market power, noting that market power is often relevant to antitrustliability but not sufficient to establish it. Beyond that, it may be more helpful to deferfurther analysis until considering specific practices in specific settings.

2.4.3. Conduct

In some situations, one may be able to infer the presence of market power from thechallenged conduct itself. If we observe a firm engaging in a practice that could not beprofitable unless it enhanced the firm’s market power to some certain degree, we maythen infer that market power would indeed increase to that degree. For example, if a firmpays large amounts to retailers to agree not to deal with prospective entrants or spendslarge sums to maintain tariffs, we may infer that these practices create or enhance thatfirm’s market power.26 If one accepts the premise that a firm’s expertise in assessing itsown market power is likely to be more reliable than that produced by a battle of expertsbefore an agency or in litigation, then the firm’s own conduct may be a sound basis forinferring the existence of market power.

25 See Carlton (2004), McAfee, Mialon, and Williams (2004), and Schmalensee (2004) for recent discussionsof how to apply the concept of entry barriers in antitrust analysis.26 As we discuss in subsection 5.4.2 in our analysis of exclusive dealing contracts with retailers, we wouldneed to rule out pro-competitive justifications, such as those based on free riding. In the case of lobbyingto erect tariff barriers, even if the conduct enhances market power, it would not violate U.S. antitrust lawsbecause petitioning government, even to restrict competition, is exempt activity under the Noerr-Penningtondoctrine.


Two caveats should be noted. First, the amount of market power that may be inferredwill sometimes not be very great. A firm with billions of dollars in sales would happilyspend millions lobbying for tariffs even if the resulting degree of market power weretrivial. On the other hand, if a firm engages in a plan of below-cost pricing that sacrificeshundreds of millions of dollars in current profits, in the absence of other explanationsone might well infer that it anticipates a substantial degree of market power, at leastsufficient to recoup its investment.

Second, the reliability of the inference depends greatly on the lack of ambiguityregarding the character of the practice under consideration. If one is certain that theconduct would only be undertaken if it could enhance the firm’s market power (tosome requisite degree), then the inference is sound. However, often it will be contestedwhether the conduct in question was designed to and will have the effect of increasingmarket power rather than constituting a benign or even beneficial practice that increaseswelfare. For example, prices below cost may be profitable because they are predatory,or because they are introductory offers that will enhance future demand for an experi-ence good, or because they stimulate the demand for other products sold by the firm ata healthy price-cost margin. If pro-competitive explanations are sufficiently plausible,no inference of market power is warranted, at least without further investigation.

Recognizing the possibility that the conduct at issue may be pro-competitive is espe-cially important given the role that market power requirements often play in antitrust,namely, as a screening device. That is, we may require a plaintiff to prove the exis-tence of market power because we do not want to subject a wide range of behavior tothe costs of antitrust scrutiny and the possibility of erroneous liability. When the con-duct that provides the basis for inferring market power is the very same conduct underscrutiny, and furthermore when the purpose and effect of such conduct is ambiguous,permitting an inference of market power from the conduct somewhat undermines thescreening function of the market power threshold. This concern may be especially greatwhen juries serve as the finders of fact.27

2.5. Market power in antitrust law

As noted, in antitrust law the notion of market power is frequently used as a screen:a firm (or group of firms) must be shown to have some level of market power as aprerequisite to considering whether the conduct in question gives rise to antitrust lia-bility. As a result, antitrust investigations and adjudications devote substantial attention

27 This concern may help to explain the Supreme Court’s decision in Spectrum Sports v. McQuillan, 506U.S. 447 (1993), where the Court held in an attempted monopolization case that the plaintiff had to meetthe market power requirement independently of proving predatory conduct. Although the holding on its faceseems illogical (if, as the plaintiff argued, it would have been irrational to have engaged in the conduct unlessthe requisite contribution to market power were present), the actual practice under consideration may wellhave appeared to the Court to be nonpredatory, so it wished to heighten the plaintiff’s required proof before itwould allow the case to be considered by the jury.


to whether or not the requisite market power exists. In rhetoric and often in reality,this legal approach of viewing market power as something either present or absent—adichotomous classification—is at odds with the technical economic notion of marketpower as a matter of degree. Because some degree of technical market power is ubiq-uitous, it is evident that the term “market power” as used in antitrust law has anothermeaning. Nevertheless, the law’s notion of market power is quite closely related to thatof economists. A legal finding of market power constitutes not merely a declaration ofthe existence of technical market power, however trivial, but rather a conclusion thatthe degree of existing or increased market power exceeds some threshold, a benchmarkthat, as we will see, varies with the type of conduct under consideration and that in mostinstances is not clearly specified.

This feature of antitrust law’s use of a market power requirement is well illustrated bythe law of monopolization. As will be elaborated in subsection 5.2, under U.S. antitrustlaw “[t]he offense of monopoly . . . has two elements: (1) the possession of monopolypower in the relevant market and (2) the willful acquisition or maintenance of that poweras distinguished from growth or development as a consequence of a superior prod-uct, business acumen, or historic accident.”28 The requirement of “monopoly power”is conclusory in that it merely signifies that degree of market power deemed minimallynecessary and also sufficient to satisfy the first element of the offense of monopoliza-tion. It is understood that this level of market power is higher than that required in otherareas of antitrust law. Notably, the market power requirement is highest in monopo-lization cases, somewhat lower in attempted monopolization cases, and lower still inhorizontal merger cases, as will be discussed in subsections 4.4.2 and 5.2.1. However,these requirements typically are not stated quantitatively, making it difficult to knowvery precisely what is the threshold in any of these areas.

In principle, the fact that market power is a matter of degree should be recognized indesigning antitrust rules. A monopolistic act that is unambiguously undesirable mightbe condemned even if the incremental impact on market power is modest, whereas forconduct that is ambiguous, with a high risk of false positives, it may be appropriate tocontemplate condemnation only when the potential effect on market power is substan-tial. If one were minimizing a loss function in which there was uncertainty about thepractices under scrutiny, and if the degree of harm conditional on the practices beingdetrimental was rising with the extent of market power, an optimal rule could be statedas entailing a market power requirement that was highly contextual.

For practical use by agencies and in adjudication, however, a more simplified formu-lation may economize on administrative costs, provide clearer guidance to parties, andreflect the limited expertise of the pertinent decision-makers. Nevertheless, some greaterflexibility may be warranted and is indeed increasingly reflected in antitrust doctrine.The early emergence of a per se rule against price-fixing, which dispenses with proofof market power, is one illustration. Another is the increasing use of intermediate lev-els of scrutiny under the rule of reason (see subsection 3.5.2) and the implicit reliance

28 United States v. Grinnell Corp., 384 U.S. 563, 570–71 (1966).


on different market power thresholds by the antitrust agencies in reviewing horizontalmergers in different industries, despite the existence of official guidelines that purportto be uniform.

In addition to differences in the magnitude of market power thresholds and whetherthere is some flexibility regarding the requisite degree of market power, there is variationacross contexts in whether the question posed concerns the extant level of market poweror the amount by which the actions under scrutiny would increase it. In a monopoliza-tion case, the standard question is often whether a firm’s past practices have improperlycreated or maintained monopoly power, so the inquiry is usually into whether significantmarket power already exists, as reflected in the previously quoted formulation. By con-trast, in examining horizontal mergers, the focus is on whether the proposed acquisitionwould significantly increase market power.29

We believe that this distinction is overstated and potentially misleading and that thecorrect inquiry should focus largely on contributions to market power. Even in thetypical monopolization case, the relevant question is how much the past practices con-tributed to the existing situation. If the contribution is large (and if the practices arenot otherwise justifiable), it seems that there should be a finding of liability even if theresulting total degree of power is not overwhelming. (In such a case, the initial levelof market power presumably will have been rather low.) Likewise, even if the degreeof existing market power is great, in cases in which the practices in question did notplausibly contribute significantly to that result, one should be cautious in condemningthose practices, that is, they should be condemned only if they are unambiguously un-desirable.

As an example, consider a firm selling a relatively homogeneous product, such as inthe chemical industry, that enjoys a significant cost advantage over its rivals based onpatented process technology. That firm might well enjoy a nontrivial degree of technicalmarket power. Neither good sense nor existing law ordinarily condemns the discoveryof a superior production process. Let us assume that the firm’s technical market powerwas legally obtained and suppose further that the firm prices against a perfectly elasticrival supply at some trigger price that is below the firm’s monopoly price. Antitrustissues could arise if this firm attempts to acquire its rivals or if the firm engages inconduct that drives its rivals out of business. In considering such cases, the degree ofthe firm’s initial market power is of secondary importance (although if it were nearzero, further inquiry would probably be pointless). Instead, the central question shouldbe whether and to what extent the acquisition or exclusionary conduct will augmentthat firm’s market power and thus harm consumers. For example, however great is theinitial level of market power, the firm would gain no additional power by acquiring(or destroying) one of its rivals as long as numerous others that remain still have highlyelastic supply at that same trigger price. However, the firm might well gain market powerby acquiring or destroying a rival with uniquely low costs, thereby raising the price at

29 See subsection 4.4.2, where we discuss the point that the extant level of market power is also important.


which substantial competing supply would be triggered. We return to the question of therelevance of extant market power versus challenged practices’ contribution to power insubsection 5.1.2 with regard to monopolization and exclusionary practices.

A further possible deviation between economic analysis and antitrust law with regardto market power concerns the benchmark against which the height of price-cost marginsis assessed. The U.S. antitrust enforcement agencies in the Horizontal Merger Guide-lines (1992) define market power as “the ability profitably to maintain prices abovecompetitive levels for a significant period of time,” and the Supreme Court has similarlystated that “As an economic matter, market power exists whenever prices can be raisedabove the levels that would be charged in a competitive market.”30 If one understandsthe competitive price to refer to the price that would be charged in a hypothetical, text-book, perfectly competitive market in which firms have constant marginal costs equalto the marginal cost of the firm in question at the prevailing equilibrium, then the legaland economic concepts are essentially the same. However, the hypothetical competitivescenario that underlies such statements is rather vague: the counterfactual is not explicit,and some specifications that may implicitly be contemplated may not yield sensible an-swers. For example, what is meant by the perfectly competitive price in a market withfixed costs?

Courts have struggled with these issues for many years. The Supreme Court has statedthat “Monopoly power is the power to control prices or exclude competition.”31 Thisis not a meaningful screen, however, since any firm with technical market power hassome ability to control prices. Conversely, in the European Union, the European Courtof Justice has said that a “dominant position” corresponds to “a position of economicstrength enjoyed by an undertaking which enables it to hinder the maintenance of ef-fective competition on the relevant market by allowing it to behave to an appreciableextent independently of its competitors and customers and ultimately of consumers.”32

This test is not especially useful either, since even a firm with great market power doesnot rationally behave independently of its competitors or customers. That is, there issome monopoly price, PM , which—however high it may be—implies that a price of,say, 2PM would be less profitable due to far greater consumer substitution away fromthe product at that higher price.

3. Collusion

We now turn to collusion, including price-fixing cartels and other arrangements that mayhave similar effects, such as the allocation of customers or territories to different suppli-

30 Jefferson Parish Hospital District No. 2 v. Hyde, 466 U.S. 2, 27 n. 46 (1984).31 U.S. v. E.I. du Pont de Nemours & Co, 351 U.S. 377, 391 (1956).32 Case 322/81, Michelin v. Commission [1983] ECR 3461 §30.


ers.33 For concreteness, we will ordinarily focus on price-fixing. There is an enormousliterature on the economics of collusion that we do not attempt to review systematically.Existing surveys include Shapiro (1989), Jacquemin and Slade (1989), Motta (2004,ch. 4), and of particular note Whinston (2006, ch. 2). For in-depth discussion of someespecially interesting price-fixing cases, see Borenstein (2004), Connor (2004), Elzingaand Mills (2004), Motta (2004, pp. 211–219), and Porter and Zona (2004).

The focus here, as in the rest of this survey, is on the intersection of economics andthe law. We begin by noting the core elements from each field and posing questionsabout their relationship. Next, we explore the economics of collusion, focusing on thenecessary elements for successful collusion, lessons from game-theoretic models ofoligopoly, and the various factors that bear on the likelihood of successful collusion.Finally, we examine legal prohibitions in light of the basic teachings of economics.

3.1. Economic and legal approaches: an introduction

3.1.1. Economic approach

For as long as there has been commercial competition, rivals have been tempted toshort-circuit it because self-interest favors their own profits at the expense of customers’interest in lower prices and the overall social interest in allocative efficiency. No lessa champion of the free-market system than Adam Smith ([1776] 1970, bk 1, ch. X)considered collusion an ever-present danger. “People of the same trade seldom meettogether, even for merriment and diversion, but the conversation ends in a conspiracyagainst the public, or in some contrivance to raise prices.” If one thinks in terms of ahomogeneous product, firms seek to establish and maintain the monopoly price, whichexceeds the price that would prevail in the absence of the agreement. With differentiatedproducts or price discrimination, although there is no single monopoly price, the sameidea applies: firms seek to elevate prices and thus raise their collective profits at theexpense of consumers. In so doing, the firms typically increase the gap between price(s)and marginal cost(s) and thus raise deadweight loss and lower total welfare, definedas the sum of supplier profits and consumer surplus. Thus, collusion is unwelcome,whether one is seeking to maximize overall efficiency or consumer welfare.

Colluding firms use a variety of methods to achieve the basic goal of raising prices. Insome cases, firms agree to minimum prices. In others, they agree to limit their produc-tion levels, since output restrictions translate into elevated prices. Alternatively, firmscan allocate customers or territories among themselves, with each firm agreeing not tocompete for customers, or in territories, assigned to others. These customer and terri-torial allocation schemes effectively grant each firm a monopoly over some portion of

33 We do not explicitly address the full range of “horizontal agreements,” which includes group boycotts aswell as arrangements among buyers, notably, to suppress the prices of inputs, the latter of which are subjectto similar analysis as that presented here, although they have received less antitrust scrutiny.


the overall market, so they lead to higher prices and reduced output, even though theseschemes do not directly specify price or output.

Economists studying collusion, and more generally oligopoly, tend to inquire intothe factors that determine the market equilibrium outcome in an industry. Economiststypically focus on whether the outcome is relatively competitive, with prices close tomarginal cost, or at least some measure of average cost, or relatively collusive, withprices close to the level that a monopolist would pick to maximize industry profits. Thisapproach is consistent with economists’ traditional emphasis on market outcomes andtheir implications for the allocation of resources.

This approach focuses on description or prediction, not on policy prescriptions re-garding how the government should mitigate the costs of collusion. There is a generalconsensus that clearly identifiable attempts to engage in collusive behavior should beprohibited, so explicit cartel agreements should not be legally enforceable and privateattempts to agree upon and enforce supra-competitive prices should be punished. It iswidely recognized, however, that it is not always possible to determine whether col-lusion is occurring or, even when it is, which specific practices should be proscribed.One approach in such settings would be price regulation, which is often undertakenin the case of natural monopolies but is generally thought to be inferior to decentral-ized competition when such is feasible. In the past, there have been recommendationsto deconcentrate certain industries in order to achieve more competitive outcomes.34

Such proposals have not been implemented, except in some cases of monopolization,and have not of late been actively considered in the United States. Another structuralapproach is more prevalent: enjoining horizontal mergers that make collusive outcomesmore likely, a topic we explore in section 4. Finally, for cases in which collusion can beidentified but the specific practices enabling it cannot, Posner (1969, 2001) interestinglyproposes the imposition of monetary penalties on oligopolists if the market equilibriumoutcome is collusive.35 The idea is that, just as Pigouvian taxes induce firms to refrainfrom inefficient behavior, the details of which might be difficult for a regulator to ob-serve or proscribe, so too would appropriate fines or damage awards in private litigationlead firms to abstain from collusive behavior. This approach assumes, importantly, thatit is possible to measure the extent to which prices exceed non-collusive levels, which

34 Legislation was introduced repeatedly in the early 1970s that would have authorized the dissolution offirms in concentrated industries that had not engaged in substantial price competition over three consecutiveyears; see, for example, S.1167, March 12, 1972. This legislation was based on the White House Task ForceReport on Antitrust Policy (1968), commonly known as the Neal Report.35 On its face, present practice appears to differ significantly from Posner’s proposal. Although the UnitedStates and most other competition enforcement regimes do provide for fines or private damage remedies incases of price-fixing, to trigger such payment obligations, the government or private parties typically need toshow that in some sense there is an “agreement.” Furthermore, satisfaction of this requirement is generallyunderstood to entail more than demonstrating that the observed outcome involves a “collusive price,” althoughas we shall discuss, just how much more must be shown and what constitutes an adequate demonstration isunclear. See subsections 3.1.2 and 3.4.


poses both conceptual and practical challenges of a sort that are encountered in impos-ing conventional price-fixing sanctions, the magnitude of which depends on the extentof collusive overcharges. However, Posner’s approach has not been embraced by thecourts.

3.1.2. Legal approach

The legal approach to collusion, at least on its face, differs from the economic approach.As just described, the economic approach begins with a diagnosis of the problem, thentries to ascertain whether and when collusion occurs, and finally assesses the efficacyof competing remedies. Although one would like to believe that the legal approach is atsome level grounded in such analysis, on the surface it appears to focus instead on par-ticular behavioral elements. As will be seen in the course of our analysis in section 3, theextent to which the legal approach can ultimately be rationalized on economic groundsdepends on how legal tests are interpreted.

In the United States, the European Union, and many other jurisdictions, the structureof legal prohibitions revolves around the distinction between unilateral and group be-havior. Unilateral behavior is circumscribed to a limited degree by anti-monopolizationlaw (see section 5) and various other provisions but is not subject to a regime of priceregulation or other forms of internal micro-management of firms’ dealings.36 Thus,firms are purportedly free to set prices and other conditions of trade.37 This freedom,however, is restricted to unilateral behavior. Independent firms are expected to compete,conferring the benefits of competition on consumers and on society as a whole.

The central legal question with which we will be concerned—and will elaborate insubsection 3.4—is how courts or other regulators are to determine when supposedlycompeting firms are instead conspiring. Legal prohibitions are typically triggered bycertain types of conduct rather than by outcomes themselves. For concreteness, wewill discuss the prohibition in U.S. antitrust law, Sherman Act §1, which makes ille-gal “[e]very contract, combination . . . , or conspiracy, in restraint of trade.” In practice,the standard term of art is “agreement,” even though that term does not appear in thestatute.38 Thus, the legal question is whether firms’ pricing is the result of an agreement.If not, there is no violation. If so, there is a violation, and penalties in the United Statesinclude having to pay treble damages to injured customers, being subject to injunctionson prohibited behavior, and criminal penalties, under which firms’ executives convictedof price-fixing serve prison terms and firms pay fines.

36 It is true, however, that remedies in monopolization cases and some others can entail what is tantamountto fairly detailed regulation.37 There are important qualifications, notably with regard to proscriptions on predatory pricing (see subsec-tion 5.3), but the focus in this section is on prices that are too high and thus directly harm customers ratherthan on prices that are too low and thus directly harm competitors.38 Interestingly, this is the language of Article 81 of the competition law in the European Union. Like in theUnited States, the concept in the European Union embraces more than formal contracts yet it is uncertain justhow much more.


What, then, is an agreement? And how does this concept relate to the economicanalysis of collusive behavior? These questions will occupy much of the remainderof section 3 of our survey. To provide some guidance in the interim, a few prelim-inary observations are offered. First, there are clear cases. At one extreme, if com-petitors meet in the proverbial smoke-filled room, negotiate a detailed cartel arrange-ment, sign it, and implement it—and, importantly, this all can be proved in a legalproceeding—an agreement and hence a legal violation will undoubtedly be found toexist. At the other extreme, no agreement would presumably exist and no violationwould be found due to the mere fact that competitors’ prices are equal—as one ex-pects with homogeneous products and perfect competition, for example—or that theysometimes move together—as tends to occur when there are shocks to input prices(think

Documents

Antitrust - Faculty Bios | Berkeley Haasfaculty.haas.berkeley.edu/shapiro/antitrust2007.pdfindustrial organization economics that are relevant for antitrust law. That task is far too